Dissertação de Mestrado de Antonio Marcio Adiodato de Menezes

Defesa de Dissertação de Mestrado Acadêmico

Aluno(a): Antonio Marcio Adiodato de Menezes
Orientador(a): Viviane Pereira Moreira

Título: Emotion Classification from the Perspective of Intratextual Entities
Linha de Pesquisa: Processamento de Linguagem Natural

Data: 16/12/2025
Hora: 13:30
Local: Esta banca ocorrerá de forma remota. Acesso público disponibilizado pelo link https://mconf.ufrgs.br/webconf/00149248.

Banca Examinadora:
-Larissa Astrogildo de Freitas (UFPEL)
-Lucas Rafael Costella Pessutto (UFRGS)
-Dennis Giovani Balreira (UFRGS)

Presidente da Banca: Viviane Pereira Moreira

Resumo: Emotions are integrated into human cognition and, naturally, manifest through written language. Their forms of expression in texts have been extensively investigated in the area of sentiment analysis. However, traditional emotion classification typically focuses on the sentence-level and on the perspective of the author, overlooking the distinct emotional viewpoints of intratextual entities (i.e., the entities that are mentioned in the text). This work proposes EmoPIE, a multi-label model for emotion classification of intratextual experiencers. EmoPIE can handle cases in which emotions are associated with multiple entities within the same text. By combining entity-specific and global contextual representations, EmoPIE enhances entity-level emotion detection. Experiments using the x-enVENT dataset have shown that EmoPIE outperformed a recent baseline across several emotion categories, demonstrating the relevance of incorporating entit y perspective into emotion modeling. The study provides a low-cost yet computationally efficient approach that may support future research in experiencer-specific emotion classification.

Palavras-Chave: Experiencer-specific Emotion Classification, Dataset Analysis.